TY - JOUR T1 - Evidence for a probabilistic, brain-distributed, recursive mechanism for decision-making JF - bioRxiv DO - 10.1101/036277 SP - 036277 AU - JA Caballero AU - MD Humphries AU - KN Gurney Y1 - 2016/01/01 UR - http://biorxiv.org/content/early/2016/02/07/036277.abstract N2 - Decision formation recruits many brain regions, but the computation they jointly perform is unknown. We introduce a recursive ideal Bayesian observer that makes choices using evidence streams with the statistics of spike trains in the macaque sensory cortex. We show it decides faster than monkeys, indicating they lose information from their sensory cortex; conversely, if equivalently deprived of information, its choice behaviour quantitatively matches that of monkeys. The algorithm's recursive architecture maps to the recurrent cortico-basal-ganglia-thalamo-cortical loops, whose components have all been implicated in decision-making. With this mapping, we show the dynamics of variables within the algorithm qualitatively match those recorded from neurons in the monkey cortex and striatum during decisions, and predict the dynamics of basal ganglia output and thalamus. Our principled, single-equation algorithm is probabilistic, distributed, recursive and parallel. Its success at capturing anatomy, behaviour and electrophysiology suggests that the mechanism approximated by the brain has these same characteristics. ER -